• Quick note - the problem with Youtube videos not embedding on the forum appears to have been fixed, thanks to ZiprHead. If you do still see problems let me know.

Intrade

BenBurch said:
Right now the market gives a 58% chance of an Obama re-election;

http://www.intrade.com/v4/markets/contract/?contractId=743474

I know these markets tend to be very accurate a few months out, but has it been studied how accurate they are 18 months out?
Click to expand...

Actually these markets do not tend to be accurate at all; in 2000 the IEM gyrated back and forth wildly on the last day, and oddly enough, the shares in Gore's stock ended up paying off as that market was based on the popular vote, not the electoral college.

In 2004, similar wild trading happened on the final day as the flawed exit poll reports resulted in a sudden swing to Kerry, and then an equally sharp and sudden shift to Bush as the actual results came in.

What InTrade and the IEM do very well, is give a snapshot in time of the conventional wisdom about the status of the campaign. Most presidents do win reelection for a second term, and thus the CW today is that Obama is more likely to win than lose in 2012. However, as has been discussed in several threads, Obama's shares have dipped from the mid-60s in the last few months. Again, this reflects the conventional wisdom.
 
The problem is, neither result would show that those odds aren't accurate.

Intrade basically reflects conventional wisdom. It was wrong about the New Hampshire primaries in 2008 for example.
 
Brainster said:
Actually these markets do not tend to be accurate at all; in 2000 the IEM gyrated back and forth wildly on the last day, and oddly enough, the shares in Gore's stock ended up paying off as that market was based on the popular vote, not the electoral college.
Click to expand...

So isn't this evidence of the accuracy of that particular market? It was a crazy close election, and Gore did win the popular vote overall.
 
While it is an interesting polling method, I don't see any real connection to reality.

Sort of like Ground Hog Punxatawny Phil predicting the length of winter, or skirt length to stock market, there is no cause/effect there.
 
Put your money down and find out. Look at the spreads in different ways of betting on Intrade.
 
casebro said:
While it is an interesting polling method, I don't see any real connection to reality.
Click to expand...

Actually prediction markets are not polling methods.

They don't care what your opinion is or who you will vote for. For example, a strong Obama opponent could still invest in Obama winning because he thinks that's what will happen even if he doesn't think that's what should happen--and even if he has no intention of voting for Obama himself.

ETA:
Sort of like Ground Hog Punxatawny Phil predicting the length of winter, or skirt length to stock market, there is no cause/effect there.
Click to expand...

I disagree. Sports betting is essentially a prediction market. There is no cause/effect (unless there's cheating), yet the "Vegas line" is a pretty accurate prediction of the expected outcome of a game especially the nearer to game time you get. The question in the OP is very appropriate. We know these kinds of predictions become less reliable the farther in advance they're made.

A prediction does not have to have a causal relation to be accurate. Think of weather forecasting. It's highly accurate within the next 24 hours, yet no one thinks meteorologists are controlling the weather somehow!
 
Last edited:
JoeTheJuggler said:
So isn't this evidence of the accuracy of that particular market? It was a crazy close election, and Gore did win the popular vote overall.
Click to expand...
If I understand the question, not necessarily.

I don't see Intrade as a "predict the winner" tool; it's simply an odds maker.

Intrade could give something a 99% chance of happening, and a failure might only mean that was the one-percent-of-the-time occurrence. You need multiple results compared with multiple predictions to see whether Intrade is accurate.
 
Ditto on the odds making and 'vegas line' comparisons.

The point of the vegas line is "everybody knows who is going to win, but we need to set the odds to get betting going." So the bet becomes "how much will Team X win by?"

I think somewhere above it was 66% for Obama? Yet I don't think there has ever been such a landslide presidential election. Not in 100 years anyhow?
So not quite reality.
 
casebro said:
Ditto on the odds making and 'vegas line' comparisons.

The point of the vegas line is "everybody knows who is going to win, but we need to set the odds to get betting going." So the bet becomes "how much will Team X win by?"

I think somewhere above it was 66% for Obama? Yet I don't think there has ever been such a landslide presidential election. Not in 100 years anyhow?
So not quite reality.
Click to expand...
Not if you mean 66% of the popular vote, in my lifetime anyway.
 
Snide said:
If I understand the question, not necessarily.

I don't see Intrade as a "predict the winner" tool; it's simply an odds maker.
Click to expand...
But it makes the odds on the predictions. (Not just winners, but all sorts of questions.) It is indeed properly called a prediction market.

Intrade could give something a 99% chance of happening, and a failure might only mean that was the one-percent-of-the-time occurrence. You need multiple results compared with multiple predictions to see whether Intrade is accurate.
Click to expand...
True. But citing an example of a prediction market accurately reflecting just what happened is certainly not evidence that "these markets do not tend to be accurate at all". That's what Brainster said in his first post to which I was responding.

Perhaps you thought I was responding in that instance to the OP. I've mostly been defending that the question the OP asks is legitimate.
 
casebro said:
Ditto on the odds making and 'vegas line' comparisons.

The point of the vegas line is "everybody knows who is going to win, but we need to set the odds to get betting going." So the bet becomes "how much will Team X win by?"
Click to expand...
The purpose of the spread is to avoid making odds. It makes all bets even money.

What that spread is can be seen as a prediction market. It's an offering, and you see how people invest. [Vegas odds makers jigger that spread in an attempt to reflect bets made at even money.]

I think somewhere above it was 66% for Obama? Yet I don't think there has ever been such a landslide presidential election. Not in 100 years anyhow?
So not quite reality.
Click to expand...
Prediction markets don't usually ask about how many electoral votes a candidate will win. As I said, they're not intended to be an opinion poll. They are strictly about predicting the winner. The really high numbers for Obama more accurately reflected the probable outcome of the election (win or loss, not number of electoral votes) than opinion polls. Opinion polls more accurately reflect popular votes.

In the 2008 election, it was very probable that Obama would win by a modest margin--especially right near election day. If the prediction market had set up a "stock" on the proposition that Obama wins by a landslide, I suspect it would have attracted few investors, and the price would have stayed very low.
 
Last edited:
JoeTheJuggler said:
True. But citing an example of a prediction market accurately reflecting just what happened is certainly not evidence that "these markets do not tend to be accurate at all". That's what Brainster said in his first post to which I was responding.

Perhaps you thought I was responding in that instance to the OP. I've mostly been defending that the question the OP asks is legitimate.
Click to expand...
Nah, I think we're in agreement pretty much. I just didn't take one data point to be evidence of anything. I would have just asked Brainster for evidence of his claim, and maybe snuck in this one anecdote as just that.
 
casebro said:
Ditto on the odds making and 'vegas line' comparisons.

The point of the vegas line is "everybody knows who is going to win, but we need to set the odds to get betting going." So the bet becomes "how much will Team X win by?"

I think somewhere above it was 66% for Obama? Yet I don't think there has ever been such a landslide presidential election. Not in 100 years anyhow?
So not quite reality.
Click to expand...

You can't read it as a comparison to the percentage of the actual vote. Let me illustrate. Suppose that one week prior to the vote, polls showed Obama leading by 20 percentage points. What would the InTrade or IEM markets show?

They would show a much greater number, because the odds of Obama winning when that far ahead in the polls would be overwhelming; they would not be 60% but maybe 95%, so that his shares might be priced at $9.95 or so. Remember, the settlement price on Obama's shares is either $10.00 or $0.00; they don't settle at the percentage of the popular vote.
 
Snide said:
Not if you mean 66% of the popular vote, in my lifetime anyway.
Click to expand...

In 1964 LBJ got 61%, which was probably the biggest landslide in the 20th century.

The electoral college vote is often much more lopsided though. In 1972 Nixon got 520 electoral votes to 17 for McGovern. The popular vote was almost as lopsided as in '64 but the electoral vote was more lopsided. Reagan did even better than that in '84 with 525 to 13.
 
Brainster said:
You can't read it as a comparison to the percentage of the actual vote. Let me illustrate. Suppose that one week prior to the vote, polls showed Obama leading by 20 percentage points. What would the InTrade or IEM markets show?

They would show a much greater number, because the odds of Obama winning when that far ahead in the polls would be overwhelming; they would not be 60% but maybe 95%, so that his shares might be priced at $9.95 or so. Remember, the settlement price on Obama's shares is either $10.00 or $0.00; they don't settle at the percentage of the popular vote.
Click to expand...

That's what I was trying to say in my most recent post.

I think that's the danger of thinking of a prediction market as an opinion poll. It's not.
 
You must log in or register to reply here.

Back
Top Bottom